Open Orchestra

Multi-agent orchestration plugin for OpenCode. Spawn, manage, and coordinate specialized AI workers from a single conversation.

Quick Install

npx open-orchestra

This interactive setup will:

1. Add the plugin to your opencode.json
2. Create .opencode/orchestrator.json with worker profiles
3. Optionally create example skill files

Manual Installation

Add to your opencode.json:

{
  "plugin": ["[email protected]"]
}

Then create .opencode/orchestrator.json:

{
  "$schema": "https://unpkg.com/[email protected]/schema/orchestrator.schema.json",
  "autoSpawn": true,
  "profiles": [
    {
      "id": "coder",
      "name": "Code Implementer",
      "model": "anthropic/claude-sonnet-4-20250514",
      "purpose": "Write and edit code"
    },
    {
      "id": "docs",
      "name": "Documentation Librarian",
      "model": "anthropic/claude-sonnet-4-20250514",
      "purpose": "Research docs and find examples",
      "supportsWeb": true
    }
  ],
  "workers": ["coder", "docs"]
}

How It Works

1. Ask the orchestrator - It analyzes your request and delegates to specialized workers
2. Workers execute - Each worker has its own model, tools, and expertise
3. Results return - Summaries flow back to the main conversation

Worker Profiles

Define workers inline in orchestrator.json or as skills in .opencode/skill/<id>/SKILL.md:

---
name: vision
description: Analyze images and screenshots
model: anthropic/claude-sonnet-4-20250514
supportsVision: true
tools:
  read: true
  write: false
---

You are a vision specialist. Describe images accurately and extract relevant details.

Features

• Multi-model orchestration - Use different models for different tasks
• Tool isolation - Workers can have restricted tool access
• Session modes - linked (shared context), child (inherited), isolated
• Memory integration - Optional Neo4j-backed knowledge graph
• Vision routing - Automatic image analysis via vision-capable workers

CLI Options

npx open-orchestra              # Interactive setup
npx open-orchestra --yes        # Accept all defaults
npx open-orchestra --minimal    # Minimal config only
npx open-orchestra --help       # Show help

Configuration Reference

See the full schema for all options.

| Field | Description | |-------|-------------| | autoSpawn | Auto-start workers on demand | | profiles | Worker definitions | | workers | Workers to pre-spawn | | memory.enabled | Enable Neo4j memory graph | | integrations | Neo4j, Linear, etc. |

License

MIT